Same difference. There's no evidence the data is biased either. The moment we start playing politics with our datasets we've screwed ourselves over.
All data is inherently biased. We don't have the luxury of working with data points in infinite dimensional space. Choices of different variables can lead to dramatically different models even with regards to the variables they have in common. And that's when we have objective qualitative data.
It feels like you are trying to argue that Uber shouldn't pay women more per job to balance out this effect, but I don't think anyone in this topic is saying they should. But it would be improper to try to use this data set to say "Men are just better at earning money than women".